Inequality constraint cause infeasible results in NMPC

I built an NMPC controller for a quarter-car system with linear state functions. The controller will give feasible results when equality constraints are specified. However, once inequality constraints were defined in a function, the controller gave infeasible results.

ValidateFcns results are OK for both constraints. Is there any problem with the definition of inequalities in the custom inequality function ?

Equality Constraints:

ManipulatedVariables(i).Min/Max= -/+5000

Inequality Constraints:

u(2)−5000≤0 / u(2)+5000≤0

cineq = [U(2) - 5000; U(2) + 5000;]

Answer 1

[Torsten]

Your constraints say that a number should be >= -5000 (Min/Max constraint) and that it should be <= -5000 (U(2)+5000). That's hard to handle for a solver. Just set U(2) = -5000 in your optimization problem.

[Mert Buyukkopru]

Thanks for the answer.

The main issue is that the constraints in my problem are not constant. Indeed they are dynamic depending on system states. Therefore, I used NMPC instead MPC to let inequality constraint in the problem.

The states x(2) and x(4) are the parameters which define the lower and upper bound at each time step.

For instance ;

cineq=myfunction(x,u,c)

if x(2)> x(4)

cineq=[300-u(2); u(2)-5000;] % 300<u(2)<5000

else if x(2)< x(4)

cineq=[u(2)+200; 4000-u(2);] % -4000<u(2)<-200

end

In fact the constant terms should be replaced with "a/b/c/d*(x(2)- x(4))" like below.

300<u(2)<5000 := a*(x(2)- x(4))<u(2)<b*(x(2)- x(4))

-4000<u(2)<-200 := c*(x(2)- x(4))<u(2)<-d*(x(2)- x(4))

[Torsten]

5000<u(2)<-5000 := a*(x(2)- x(4))<u(2)<-a*(x(2)- x(4))

Do you know a number that is greater than 5000 and smaller than -5000 ? I don't.

[Mert Buyukkopru]

Yes, you are right. I modified the inequalities with the ones I use. But there are still some problem.